extension field cancellation
play

Extension Field Cancellation A New MQ Trapdoor Construction - PowerPoint PPT Presentation

Extension Field Cancellation A New MQ Trapdoor Construction February 2016 Alan Szepieniec 1 , Jintai Ding 2 , Bart Preneel 1 1: KU Leuven, ESAT/COSIC first.secondname@esat.kuleuven.be 2: University of Cincinnati, jintai.ding@uc.edu 1/24


  1. Extension Field Cancellation A New MQ Trapdoor Construction February 2016 Alan Szepieniec 1 , Jintai Ding 2 , Bart Preneel 1 1: KU Leuven, ESAT/COSIC first.secondname@esat.kuleuven.be 2: University of Cincinnati, jintai.ding@uc.edu 1/24

  2. Outline • Introduction • Extension Field Cancellation • Basic Trapdoor • Frobenius Tail • Attacks and Defenses • Bilinear Attack • Algebraic Attack – Minus • Differential Symmetry – Projection • Security & Efficiency • Security Estimation • Implementation Results • Conclusion 2/24

  3. Multivariate Quadratic Cryptosystems • public key: P ∈ ( F q [ x 1 , . . . , x n ]) m • public operation: evaluate in x ∈ F n q • secret key: ( S, T, F ) where S ∈ GL n ( F q ) , T ∈ GL m ( F q ) , F ∈ ( F q [ x 1 , . . . , x n ]) m such that P = T ◦ F ◦ S • private operation: invert S, F , T — all easy! encryption or signature verification P public knowledge private knowledge S F T decryption or signature generation 3/24

  4. Single-Field Schemes • all arithmetic occurs in F q • canonical example: UOV   � o � � o � o T v T � o T v T � � � • F i ( o , v ) = = F i   v v • invert F ( o , v ) = y : • fix v at random • solve F ( o , v ) = y for o • linear system! 4/24

  5. Mixed-Field Schemes • arithmetic occurs in F q as well as in F q n ∼ = F q [ z ] / � p ( z ) � • canonical example: HFE • let ϕ ( x ) : F n q → F q n : x �→ X = x 0 + x 1 z + . . . x n − 1 z n − 1 j<d α i,j X q i + q j + � k<d β k X q k + γ • let f ( X ) = � � i<d • F ( x ) = ϕ − 1 ◦ f ◦ ϕ ( x ) • or for simplicity: F ( X ) = f ( X ) • invert F ( X ) = Y : • factorize the polynomial F ( X ) − Y • choose a root X r such that F ( X r ) − Y = 0 5/24

  6. MQ Encryption Schemes • ZHFE • mixed-field • 2 high-degree polynomials F ( X ) and ˆ F ( X ) linked to 1 low-degree polynomial Ψ( X ) • inversion: factorize Ψ( X ) • ABC / Simple Matrix Encryption • single-field, but embeds matrix algebra • reduces inversion to linear system solving • Extension Field Cancellation (EFC) • mixed-field • 2 high-degree polynomials • reduces inversion to linear system solving !! All three are expanding maps F n q → F 2 n !! q 6/24

  7. EFC: Basic Trapdoor • let ϕ m : F n q → F n × n map a vector x ∈ F n q to the matrix q representation of X ∈ F q n . • let A, B ∈ F n × n be matrices and q α ( X ) = ϕ ( A x ) , β ( X ) = ϕ ( B x ) • Central map: � ϕ m ( A x ) x � � α ( X ) X � F = = ϕ m ( B x ) x β ( X ) X 7/24

  8. EFC: Basic Trapdoor Central map: � ϕ m ( A x ) x � � α ( X ) X � F = = ϕ m ( B x ) x β ( X ) X How to invert? � α ( X ) X � � D 1 � F ( X ) = = β ( X ) X D 2 Solution: β ( X ) D 1 − α ( X ) D 2 = 0 i.e. , solve for x : ϕ m ( B x ) d 1 − ϕ m ( A x ) d 2 = 0 which is a linear system. 8/24

  9. Enhanced Trapdoor • key idea: use Frobenius isomorphism • disadvantage: restricted to characteristic 2 only � α ( X ) X + β ( X ) 3 � E ( X ) = β ( X ) X + α ( X ) 3 9/24

  10. Enhanced Trapdoor: Inversion How to invert? � α ( X ) X + β ( X ) 3 � � D 1 � E ( X ) = = β ( X ) X + α ( X ) 3 D 2 Solution: solve for X : α ( X ) D 2 − β ( X ) D 1 = α ( X ) 4 − β ( X ) 4 or for x : α m ( x ) d 2 − β m ( x ) d 1 = Q 2 ( A x − B x ) where Q 2 ∈ F n × n is the matrix associated with the Frobenius q transform X �→ X 4 . 10/24

  11. Bilinear Attack � α ( X ) X � � Y 1 � • basic variant: F ( X ) = = β ( X ) X Y 2 • bilinear relation: β ( X ) Y 1 = α ( X ) Y 2 • there exists coefficients K i , L i ∈ F q n such that n − 1 X q i ( K i Y 1 + L i Y 2 ) = 0 � i =0 • attack: • generate many tuples ( X , Y 1 , Y 2 ) • compute K i and L i using linear algebra • given a ciphertext Y = ( Y 1 , Y 2 ) and given the coefficients K i , L i , computing X is easy 11/24

  12. Other Attacks and Defenses • same basic idea • protect against Bilinear Attack: minus • protect against Algebraic Attack: more minus • protect against Differential Symmetry Attack: projection • EFC − p , EFC − pt 2 12/24

  13. Algebraic Attack • Algebraic Attack: decent Gr¨ obner bases algorithms ( e.g. F 4 , F 5 , MutantXL) • Running time depends on degree of regularity • D reg depends on rank of quadratic form � X T F 1 X � X T = ( X , X q , X q 2 . . . X q n − 1 ) F ( X ) = where e.g. X T F 2 X 13/24

  14. Rank of Extension Field Quadratic Form F 1 = α ( X ) X ∼ F ◦ S ∼ rank = 2 rank = 2 (change of basis) T ◦ F ◦ S ∼ full rank T ( X ) = � t i X q i T ◦ F ( X ) = � t i � q i X T F X � 14/24

  15. Fast Gr¨ obner Basis F 4 • F 4 implicitly recovers T 15/24

  16. Minus • solution: drop a rows from T • F 4 can only recover n − a rows of T F 4 • rank r = 2 + a • drawback: guess a values during decryption 16/24

  17. Effect of Minus • fixed n = 35 65536 16384 4096 1024 time 256 64 16 4 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 number of applications 17/24

  18. Decryption Errors 1 a = 10 a = 12 a = 8 0.8 a = 6 0.6 error rate a = 4 0.4 a = 2 0.2 a = 0 0 1 5 10 15 20 25 n 18/24

  19. Differential Symmetry Attack • D F ( x , y ) = F ( x + y ) − F ( x ) − F ( y ) + F ( 0 ) • symmetry ⇔ ∃ Λ , L . D F ( L x , y ) + D F ( x , L y ) = Λ D F ( x , y ) • broke SFLASH • solution (pSFLASH): S must be singular and n prime • EFC p : • rank ( A ) = rank ( B ) = n − 1 • n is prime • and ker ( A ) ∩ ker ( B ) = { 0 } 19/24

  20. Estimating Security • algebraic attack: Gaussian elimination in matrix with � n � T = monomials D reg � n � • τ = nonzero terms per row 2 • complexity of Wiedemann algorithm: O ( τT 2 ) • D reg ≤ ( q − 1)( r + a ) + 2 2 t 2 n q a D reg security 83 2 10 8 82 83 2 8 8 82 59 3 6 10 82 20/24

  21. Decryption Time as a Function of a 256 64 decryption time (seconds) 16 4 1 0.25 0.0625 0.015625 0.00390625 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 a 21/24

  22. Algebraic Attack Time • implementation in Magma (has F 4 ) 65536 EFC − p , a = 10 16384 EFC − pt 2 , a = 8 4096 1024 256 64 time 16 4 1 0.25 0.0625 0.015625 15 20 25 30 35 38 n 22/24

  23. Implementation Results construction sec. key pub. key ctxt. EFC − p , q = 2 , n = 83 , a = 10 48.3 KB 509 KB 20 B EFC − pt 2 , q = 2 , n = 83 , a = 8 48.3 KB 523 KB 20 B EFC − p , q = 3 , n = 59 , a = 6 48.8 KB 375 KB 28 B construction key gen. enc. dec. EFC − p , q = 2 , n = 83 , a = 10 2.45 s 0.004 s 9.074 s EFC − pt 2 , q = 2 , n = 83 , a = 8 3.982 s 0.004 s 2.481 s EFC − p , q = 3 , n = 59 , a = 6 2.938 s 0.004 s 12.359 s 23/24

  24. Conclusion • extension field cancellation (EFC) • MQ mixed field trapdoor construction • generate a pair of high-degree quadratic polynomials • uses commutativity of extension field to cancel the polynomials’ complexity • end up with a linear system • modifiers • Frobenius Tail in char 2 (speed) • Minus (protects against Algebraic Attack) • Projection (destroys Differential Symmetry) • future work • get rid of Minus modifier • better security argument • shrink public keys • hardware implementation 24/24

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend